Improving frequency estimation of fingerprint minutia configurations using automated pre-selection

Likelihood ratio based statistical reporting on the comparison of fingerprints is thought to convey more information than traditional discrete reports such as a yes or no answer. Additionally, using a weight for each minutia based on their rarity improves the precision of the comparison process. During this research a tool was designed that assists in estimating the rarity of configurations of one or two minutiae with a core or delta as a reference position. It performs a pre-selection of fingerprints from a database by utilizing an existing minutia extraction and encoding tool and using an orientation and pixel position based matching technique. The found fingerprints are expected to be further examined manually, but since part of the database is automatically discarded the workload is reduced. The minimum requirements set are shown to be marginally reached, after which the shortcomings of the implementation are analyzed, and additional and alternative techniques are introduced that are expected to improve the tool’s performance.